Ada 2005 & IEEE 754r ?

Ada 2005 & IEEE 754r ?

[britt]

Is anyone involved with the Ada 2005 revision also tracking the
proposed changed to IEEE-STD-754 (the binary floating point arithmetic
standard)? I'm asking because the draft of 754r cites several languages
(C99, Java,C Fortran, PL/1, COBOL) in its rationale but it doesn't
mention Ada anywhere. The revised standard defines new decimal
arithmetic types and it would be a bad thing if they are somehow
incompatable with decimal types in Ada. I don't know enough to evaluate
this myself so I'm hoping someone else can look at this and see if
there are any potential conflicts.

See http://754r.ucbtest.org/drafts/754r.pdf and
http://754r.ucbtest.org/ for more information.

Britt

Re: Ada 2005 & IEEE 754r ?

[Martin Dowie]

britt@acm.org wrote:
> Is anyone involved with the Ada 2005 revision also tracking the
> proposed changed to IEEE-STD-754 (the binary floating point arithmetic
> standard)? I'm asking because the draft of 754r cites several languages
> (C99, Java,C Fortran, PL/1, COBOL) in its rationale but it doesn't
> mention Ada anywhere. The revised standard defines new decimal
> arithmetic types and it would be a bad thing if they are somehow
> incompatable with decimal types in Ada. I don't know enough to evaluate
> this myself so I'm hoping someone else can look at this and see if
> there are any potential conflicts.
> 
> See http://754r.ucbtest.org/drafts/754r.pdf and
> http://754r.ucbtest.org/ for more information.

From:
Minutes of the 24th ARG Meeting, 17-19 September 2004,
Madison, Wisconsin, USA

"AI-315/02 Full support for IEC 559:1989 (Other AI versions)

  Pascal (the AI�s author) proposes No Action. There is a big
  implementation impact, and it�s unclear that there is sufficient
  demand. Tucker wonders if we should encourage implementations to follow
  this model. Pascal says that there are problems with it; being able to
  query flags has a significant negative impact on instruction
  scheduling. So you�d have to do operations in order, which would negate
  11.6. Tucker says this still a good place to start. Steve says that
  doing this would require a non-standard mode for an implementation.

  Erhard would like to leave it open for the future, as this is a good
  idea that we couldn�t finish satisfactorily. Some AIs have been around
  for a long time (e.g. AI-51). Pascal would not like to leave it open;
  he prefers a clean list of AIs. Other AIs (like preconditions) would be
  also fall into the category of good ideas that we didn�t have
  sufficient time/effort/experience with.

  John volunteers to put information about AIs that might be good ideas
  but we couldn�t finish into the Rationale.

  No Action: 8-0-1."

Cheers

-- Martin